Church - A Set of Postulates For The Foundation of Logic

Annals of Mathematics
A Set of Postulates for the Foundation of Logic

Author(s): Alonzo Church
Source: Annals of Mathematics, Second Series, Vol. 33, No. 2 (Apr., 1932), pp. 346-366
Published by: Annals of Mathematics
Stable URL: http://www.jstor.org/stable/1968337
Accessed: 03-12-2015 16:23 UTC
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact support@jstor.org.
Annals of Mathematics is collaborating with JSTOR to digitize, preserve and extend access to Annals of Mathematics.
http://www.jstor.org
This content downloaded from 200.7.141.5 on Thu, 03 Dec 2015 16:23:47 UTC
All use subject to JSTOR Terms and Conditions
A SET
OF
POSTULATES
OF
FOR THE
LOGIC.'
FOUNDATION
BY ALONZO CHURCH.2
1. Introduction. In this paper we present a set of postulates for the

foundation of formal logic, in which we avoid use of the free, or real,
variable, and in which we introduce a certain restrictionon the law of
excluded middle as a means of avoiding the paradoxes connectedwith the
mathematicsof the transfinite.
Our reason for avoiding use of the free variable is that we requirethat
every combination of symbols belonging to our system, if it represents
a proposition at all, shall represent a particular proposition,unambigouously, and without the addition of verbal explanations. That the use of
the free variable involves violation of this requirement,we believe is
readily seen. For example, the identity
(1)
a(b+c)
ab+ac
in which a, b, and c are used as freevariables, does note state a definite

proposition unless it is known what values- may be taken on by these
variables, and this information,if not impliedin the context, mustbe given
by a verbal addition. The range allowed to the variables a, b, and c
might consist of all real numbers,or of all complex numbers,or of some
other set, or the ranges allowed to the variables might differ,and for
each possibilityequation (1) has a differentmeaning. Clearly, when this
equation is writtenalone, the propositionintendedhas not been completely
translated into symbolic language, and, in order to make the translation
complete, the necessary verbal addition must be expressed by means of
the symbols of formallogic and included, with the equation, in the formula
used to representthe proposition. When this is done we obtain, say,
(2)
R(a) R(b)J (c) Dabc . a(b + c) =
a b+ ac
where R (x) has the meaning, "x is a real number," and the symbol9)abc
has the meaning described in ?? 5 and 6 below. And in this expression
there are no free variables.
IReceived October 5, 1931.
This paper contains, in revised form,the work of the author while a National Research
Fellow in 1928-29.
346
2
POSTULATES
347
FOR THE FOUNDATION OF LOGIC.
A furtherobjection to the use of the free variable is contained in the

duplication of symbolismwhich arises when the free, or real, variable
and the bound, or apparent, variable are used side by side.3 Corresponding
to the proposition, represented by equation (1) when a, b, and c stand
for any three real numbers,there is also a propositionexpressed without
the use of free variables, namely(2), and between these two propositions
we know of no convincing distinction. An attempt to identifythe two
propositionsis, however, unsatisfactory,because substitutionof (1) for(2),
when the latter occurs as a part of a more complicatedexpression, cannot
always be allowed without producing confusion. In fact, the onlyfeasible
solution seems to be the complete abandonmentof the free variable as
a part of the symbolismof formal logic.'
Rather than adopt the method of Russell for avoiding the familiarparadoxes of mathematicallogic,5 or that of Zermelo,6 both of which appear
somewhatartificial,we introducefor this purpose,as we have said, a certain
restrictionon the law of excluded middle. This restriction consists in
leaving open the possibilitythat a propositionalfunctionF may, for some
values X of the independentvariable, representneither a true proposition
nor a false proposition. For such a value X of the independentvariable
we suppose that-F (X) is undefinedand represents nothing, and we use
a system of logical symbols capable of dealing with propositionalfunctions
whose ranges of definitionare limited.
In the case of the Russell paradox the relevance of this proposed restriction on the law of excluded middleis evident. The formulaP which leads
to this paradox may be written, in the notation explained below,
) .,--op(gp)}
I(I -. So(Sp)). It has the propertythat if we assume P then
we can inferP and if we assume P then we can infer P. On ordinary
assumptionsboth the truthand the falsehood of P can be deduced in consequence of this property,but the system of this paper, while it provides
for the existence of a propositionalfunction S . Sp(p) does not provide
either that this propositionalfunctionshall be true or that it shall be false,
for the value I S-!. So(S) of the independentvariable.
Other paradoxes either disappear in the same way, or else, as in the
case of the Epimenides or the paradox of the least undefinableordinal,
theintroductionto the second editionofWhitehead and Russell's Principia Mathematica.
4Unless it is retained as a mere abbreviation of notation.
5 B. Russell, Mathematical Logic as based on the Theory of Types, Amer. Jour.Math.,
vol. 30 (1908), pp. 222-262. A list of some of these paradoxes, with a referenceto the
source of each, will be found in this article, or in Whitehead and Russell, Principia
Mathematica, vol. 1, pp. 63-64.
6E. Zermelo, Untersuchungen uber die Grundlagen der Mengenlehre, Math. Annalen,
vol. 65 (1908), pp. 261-281.
3 Cf.
23*
348
A. CHURCH.
in termsof the undefined

theycontainwordswhichare not definable
us.
and henceneednotconcern
of oursystem,
symbols
so readilydisposedof.
is not,however,
The paradoxof Burali-Forti
or
ofourpostulates,
thisparadoxis a consequence
The questionwhether
of themwill enableus to avoidit, probablymustbe
whatmodification
whichresultsfromthe
leftopenuntilthe theoryof ordinalnumbers
has beendeveloped.
postulates
the systemof logic whichresultsfromour postulatesis
Whether
and whetherit is wholly
of mathematics,
adequateforthe development
are questions
whichwe cannotnowanswerexcept
freefromcontradiction,
answerto
by conjecture.Ourproposalis to seek at least an empirical
of the conout in somedetaila derivation
by carrying
thesequestions
and it is hopedeitherthat the systemwill
sequencesof ourpostulates,
fromcontraof adequacyand freedom
the conditions
turnoutto satisfy
or additions.
dictionor thatit can be madeto do so by modifications
2. Relation to intuitionism. Since,in the postulateset whichis
givenbelow,the law of the excludedmiddleis replacedby weaker
the questionariseswhattherelationis betweenthesystem
assumptions,
ofL. E. J.Brouwer.7
from
thisset,andtheintuitionism
oflogicwhichresults
are not the same,because,althoughbothrejecta
The two systems
of the excludedmiddle,the partsrejected
certainpartof the principle
deniedbyBrouwer,
is preserved
Thelaw ofdoublenegation,
aredifferent.
whichBrouweraccepts,
in the systemof thispaper,and the principle,
can be inferred
is false,we
from
whicha contradiction
thata statements
cases.
in
to
abandon
certain
findit necessary
to have the property,
whichrelatesit
Oursystem
appears,however,
of the formE x . F (x) (read,"there
that a statement
to intuitionism,
M
unlessthereexistsa formula
existsx suchthatF (x)") is neverprovable
suchthatF (M) is provable.
3. The abstract character of formallogic. We do not attach
or absolutetruthto any particular
system
of uniqueness
anycharacter
because
invented
logic are abstractions,
of logic. The entitiesof formal
andsystematizing
factsofexperience
orobservation,
oftheiruseindescribing
in
determined roughoutlineby thisintended
use,
and theirproperties,
on thearbitrary
choiceof theinventor.
dependfortheirexactcharacter
7 See L. E. J. Brouwer,Intuitionistische
Jahresbericht
der D. Math.Ver.,
Mengenlehre,
fur
vol. 28 (1919), pp. 203-208, and Mathematik,Wissenschaftund Sprache,Monatshefte
Math. u. Phys.,vol.36 (1929), pp. 153-164,and manyotherpapers.
use theword,"statement",
becausewe wishto reservetheword,"propo8 We purposely
in theformofa proposition,
eithertrue or false. A statement,
which
sition",forsomething
fails to be eithertrue or false,we regardas a meregroupofsymbols,withoutsignificance.
POSTULATES
349
We may draw the analogy of a three dimensional geometry used in

describingphysical space, a case for which, we believe, the presence of
such a situation is more commonly recognized. The entities of the
geometryare clearly of abstract character, numberingas they do planes
without thickness and points which cover no area in the plane, point sets
containingan infinitudeof points, lines of infinitelength, and other things
which cannot be reproduced in any physical experiment. Nevertheless
the geometry can be applied to physical space in such a way that an
extremelyuseful correspondence is set up between the theorems of the
geometryand observable facts about materialbodies in space. In building
the geometry,the proposed application to physical space serves as a rough
guide in determiningwhat properties the abstract entities shall have, but
does not assign these properties completely. Consequentlythere may be,
and actually are, more than one geometry whose use is feasible in
describingphysical space. Similarly,there exist, undoubtedly,more than
one formal system whose use as a logic is feasible, and of these systems
one may be more pleasing or more convenientthan another, but it cannot
be said that one is right and the other wrong.
In consequence of this abstract character of the system which we are
about to formulate,it is not admissible, in proving theoremsof the system,
to make use of the meaning of any of the symbols,althoughin the application which is intended the symbols do acquire meanings. The initial
set of postulates must of themselves definethe system as a formal structure, and in developing this formal structure reference to the proposed
application must be held irrelevant. There may, indeed, be other applications of the system than its use as a logic.
4. Intuitive logic. It is clear, however, that formulas composed of
symbols to which no meaning is attached cannot define a procedure of
proof or justify an inferencefrom one formulato another. If our postulates were expressed wholly by means of the symbols of formallogic
without use of any words or symbols having a meaning, there would be
no theorems except the postulates themselves. We are therefore obliged
to use in some at least of our postulates other symbols than the undefinedterms of the formal system, and to presuppose a knowledge of the
meaning of these symbols, as well as to assume an understandingof a
certain body of principles which these symbols are used to express, and
which belong to what we shall call intuitive logic.9 It seems desirable
9 The principles of intuitive logic which we assume initially form,of course, a part of
the body of facts to which the formal system, when completed, is to be applied. We
should not, however, allow this to confuse us as to the clear cut distinction between intuitive logic and formal logic.
350
A. CHURCH.
to make these presuppositionsas few and as simple as we can, but there

is no possibility of doing without them.
Before proceeding to the statement of our postulates, we shall attempt
to make a list of these principles of intuitivelogic which we find it necessary to assume and of the symbols a knowledge of whose meanings
we presuppose. The latter belong to what we shall call the language of
intuitivelogic, as distinguishedfromthe language of formal logic which
is made up of the undefinedterms of our abstract system.
We assume that we know the meaningof the words symboland formula
(by the word formulawe mean a set of symbols arranged in an order of
succession, one after the other). We assume the ability to write symbols
and to arrange them in a certain order on a page, and the ability to
recognize differentoccurrences of the same symbol and to distinguish
between such a double occurrence of a symbol and the occurrence of
distinctsymbols. And we assume the possibility of dealing with a formula
as a unit, of copying it at any desired point, and of recognizing other
formulasas being the same or distinct.
We assume that we know what it means to say that a certain symbol
or formulaoccursin a given formula,and also that we are able to pick
out and discuss a particular occurrenceof one formulain another.
We assume an understandingof the operation of substitutinga given
symbol or formula for a particular occurrenceof a given symbol or
formula.
And we assume also an understandingof the operation of substitution
throughout a given formula, and this operation we indicate by an
S, SxU j representingthe formulawhich results when we operate on the
formulaU by replacing X by Y throughout,where Y may be any symbol
or formulabut X must be a single symbol,not a combinationof several
symbols.
We assume that we know how to recognize a given formula as being
obtainable fromthe formulas of a certain set by repeated combinationsof
the latter according to a given law. This assumptionis used below in
definingthe term "well-formed". It may be described as an assumption
of the ability to make a definitionby induction,when dealing with groups
of symbols.
We assume the ability to make the assertion that a given formulais
we are constructing,
one of those belonging to the abstract systeim-which
and this assertion we indicate by the words is true. As an abbreviation,
however, we shall usually omit the words is true, the mere placing of
the formulain an isolated position being taken as a sufficientindication
of them.
POSTULATES
351
We assume furtherthe meaning and use of the word everyas part of

the language of intuitivelogic, and the use in connectionwith it of variable
letters, which we write in bold face type to distinguishthemfromvariable
letters used in the language of formallogic. These variable letters, written
in bold face type, stand always for a variable (or undetermined)symbol
Qr formula.
We assume the meaning and use of the following words from the
then,not,and is in
language of intuitive logic: thereis, and, or, if
what combinations
know
we
that
assume
we
That
is,
the sense of identity.
propositions,
constitute
other
symbols
of these words with themselvesand our
and, in a simple sense, what such propositionsmean.
We assume that we know how to distinguishbetween the words and
symbols we have been enumerating,which we shall describe as symbols
of intuitive logic, and other symbols, which are mere symbols, without
meaning, and which we shall describe as formal symbols.
We assume that we know what it is to be a proposition of intuitive
logic, and that we are able to assert such propositions, not merely one
proposition,but various propositionsin succession. And, finally,we assume
the permanency of a proposition once asserted, so that it may at any
later stage be reverted to and used as if just asserted.
In making the preceding statements it becomes clear that a certain
circle is unavoidable in that we are unable to make our explanations of
the ideas in question intelligible to any but those who already understand
at least a part of these ideas. For this reason we are compelled to
assume them as known in the beginningindependentlyof our statement
of them. Our purpose has been, not to explain or convey these ideas,
but to point out to those who already understandthem what the ideas
are to which we are referringand to explain our symbolismfor them.
5. Undefined terms. We are now ready to set down a list of the
undefinedterms of our formallogic. They are as follows:
...
I{
()2
Al]
n,
XM2
&,
aJ
a,2 A.
The expressions { } ( ) and i [ ] are not, of course, single symbols, but

sets of several symbols, which, however, in every formulawhich will be
provable as a consequence of our postulates, always occur in groups in
the order here given, with other symbolsor formulasbetween as indicated
by the blank spaces.
In addition to the undefinedterms just set down, we allow the use, in
the formulas belonging to the system which we are constructing,of any
other formal symbol, and these additional symbols used in our formulas
we call variables.
352
A. CHURCH.
An occurrence of a variable x in a given formulais called an occurrence

of x as a bound variable in the given formulaif it is an occurrenceof x
in a part of the formulaof the formAx [M]; that is, if there is a formulaM
such that 1x[M] occurs in the given formula and the occurrence of x in
question is an occurrence in Ax [M]. All other occurrences of a variable
in a formulaare called occurrences as a free variable.
A formula is said to be well-formedif it is a variable, or if it is one
of the symbolsH, A, &, a-, &, A, or if it is obtainable fromthese symbols
by repeated combinationsof them of one of the forms{M} (N) and I x [Ml,
where x is any variable and M and N are symbols or formulaswhich are
being combined. This is a definitionby induction. It impliesthe following
rules: (1) a variable is well-formed(2) H, X, &, a1%-, &, and A are well-formed
(3) if M and N are well-formedthen {M} (N) is well-formed(4) if x is
a variable and M is well-formedthen ) x [MI is well-formed.
All the formulaswhich will be provable as consequences of our postulates
will be well-formedand will contain no free variables.
The undefinedterms of a formal system have, as we have explained,
no meaningexcept in connectionwith a particularapplication of the system.
But for the formal system which we are engaged in constructingwe have
in mind a particular application, which constitutes,in fact, the motivefor
constructingit, and we give here the meanings which our undefinedterms
are to have in this intended application.
If F is a functionand A is a value of the independentvariable forwhich
the functionis defined,then {F} (A) representsthe value taken on by the
functionF when the independentvariable takes on the value A. The usual
notation is F(A). We introduce the braces on account of the possibility
that F might be a combinationof several symbols,but, in the case that F
is a single symbol,we shall oftenuse the notationF(A) as an abbreviation
for the fuller expression.10
Adopting a device due to Schonfinkel," we treat a function of two
variables as a functionof one variable whose values are functionsof one
variable, and a functionof three or more variables similarly. Thus, what
is usually writtenF (A,B) we write {{F} (A)} (B), and what is usually written
F(A, B, C) we write {{{F} (A)} (B)} (C), and so on. But again we frequently
findit convenientto employ the more usual notations as abbreviations.
If M is any formulacontainingthe variable x, then 2 x [M] is a symbol
for the functionwhose values are those given by the formula. That is,
"'The braces { } are, as a matter of fact, superfluous and might have been omitted
from our list of undefinedterms, but their inclusion makes for readability of formulas.
11M. Schonfinkel,lYber die Bausteine der mathematischenLogik, Math. Annalen, vol. 92
(1924), pp. 305-316.
POSTULATES
353
I x[M] represents a function,whose value for a value L of the independent

variable is equal to the result Sx MI of substitutingL forx throughoutM,
whenever Sx MI turns out to have a meaning, and whose value is in any
other case undefined.
The symbol 17 stands for a certain propositional functionof two independent variables, such that H(F, G) denotes, " G(x) is a true proposition
for all values of x for which F (x) is a true proposition." It is necessary
to distinguish between the proposition H (F, G) and the proposition
_x F(x)D G(x) (read, "For every x, F(x) implies G(x)").
The latter
proposition justifies, for any value M of x, the inference F(M) 9 G(M),
and hence can be used only in the case that the functionsF and G are
definedfor all values of their respective independentvariables. The propositionH(F, G) does not, on the otherhand, justifythis inference,although,
when {F}I(M) is known to be true, it does justify the inference{G}(M).
And the propositionH(F, G) is, therefore,suitable for use in the case that
the ranges of definitionof the functionsF and G are limited.
The symbol X stands for a certain propositional functionof one independentvariable, such that E(F) denotes, "There exists at least one value
of x for which F(x) is true."
The symbol & stands for a certain propositional functionof two independent variables, such that, if P and a are propositions,& (P, Q) is the
logical product P-and-Q.
The symbol
stands for a certain propositional functionof one independentvariable, such that, if P is a proposition,then - (P) is the negation
of P and may be read, "Not-P".
The symbol t stands for a certain functionof one independentvariable,
such that, if F is a propositional function of one independentvariable,
then t(F) denotes, "The object x such that {F}(x) is true."
The symbolA stands for a certainfunctionof two independentvariables,
the formulaA(F, M) being read, "The abstraction from M with respect
to F."
6. Abbreviations and definitions. In practice we do not use actually the notation just described, but introduce various abbreviations and
substituted notations, partly for the purpose of shorteningour formulas
and partly in order to render them more readable. We do not, however,
regard these abbreviations as an essential part of our theory but rather
as extraneous. When we use them we do not literally carry out the
developmentof our system, but we do indicate in full detail how this
developmentcan be carried out, and this is for our purpose sufficient.
As has been said above, we use {F} (A, B) as an abbreviation for
{{F}(A)}(B) and similarlyin the case of functions of larger numbers of
A. CHURCH.
354
variables. Moreover, alike in the case of functions of one variable and

in the case of functions of two or more variables, we omit the braces
{ I whenever the functionis representedby a single symbol rather than
by an expression consisting of several symbols. Thus, if F is a single
symbol,we write F(A) instead of {F} (A) and F(A, B) instead of {F} (A, B)
or {{F} (A)} (B).
We shall usually write [M][N] instead of & (M, N) or { {&} (M)} (N), and
--[M] instead of -(M) or {s} (M).
When F is a single symbol,we shall oftenwrite F x [M] instead of

F (2 x [MI).
Instead of H(2x[M], 2x[N]), we shall write [M]cx[N]. And
[iXy[M]]Dx[[M1Dy[N]]we abbreviate furtherto [MDOxy

[N], and
z
to [M]OXy[N] and
we
abbreviate
z
[t y[. [M]]] [[.7 [M]J y[[M]Oz[N]]]
so on.
cx
Moreover,wheneverpossible withoutambiguity,we omitsquare brackets

[1, whetherthe brackets belong to the undefinedterm2 [ ] or whetherthey
appear as a part of one of our abbreviations. In order to allow the
omission of square brackets as often as possible, we adopt the convention
that, whenever there are more than one possibility, the extent of the
omittedsquare brackets shall be taken as the shortest. And when the
omission of the square brackets is not possible without ambiguity, we
can sometimessubstitutefor them a dot, or period. This dot, when it
occurs within a parenthesis,enclosed by eithersquare brackets [ ], round
parentheses ( ), or braces { }, stands for square brackets extendingfrom
the place where the dot occurs and up to the end of the parenthesis,
or, if the parenthesis is divided into sections by commas as in the case
of functions of two or more variables, extending from the place where
the dot occurs and up to the firstof these commas or to the end of the
parenthesis,whichever is firstreached. And when the dot is not within
any parenthesis, it stands for square brackets extendingfromthe place
where it occurs and up to the end of the entireformula. In otherwords,
a dot representssquare brackets extendingthe greatest possible distance
forwardfromthe point where it occurs.
In addition to these abbreviations, we allow afreely the introduction
12 and which
of abbreviations of a simpler sort, which we call definitions,
consist in the substitutionof a particular single symbol for a particular
well-formedformulacontainingno free variables.
12 There seems to be, as a matter of fact, no serious objection to treating definitions
as an essential part of the system rather than as extraneous, but we believe it more consistent to class them with our other and more complicated abbreviations.
POSTULATES
We introduce at once the followingdefinitions,using an arrow

mean "Stands for", or, "Is an abbreviation for":
v
V rP)*
.
V
Q
E i
Si
whr
H
V) . t(V,
V nip,
y
Sp.S (7r)
355
to
10
V(P, Q) is to be read, "P or Q", U(P, Q) is to be read, "P implies Q",
Q (F, G) is to be read "F and G are equivalent" and E(M) is to be read
"M exists"
And in connection with these symbols just defined we introduce some
furtherabbreviations. For V(P, Q) we write [P] v [Q], and for U(P, Q)
we write [P19 [Q]. We abbreviate Q (Qx . M, i x . N) to [Ml x [N]. And
we abbreviate E(x) Ox [M] to 'x[M], which may be read, "For every x, M".
The notion of a class may be introducedby means of the definition:
K
A(Q).
The formulaK(F) is then to be read, "the class of x's such that (F} (x)
true."
We divide our postulates into two groups, of which
7. Postulates.
the firstconsists of what we shall call rules of procedureand the second of
what we shall call formalpostulates. The latter assert that a given formula
is true, and contain nothingfromthe language of intuitivelogic otherthan
the words is true (and even these words, as already explained, we leave
unexpressed when we write the postulates). And the former,the rules of
procedure,contain other words fromthe language of intuitivelogic.
The theoremswhich are proved as consequences of these postulates ale
of the same formas the postulates of the firstgroup, namely, that a certain
formulais true. And the proofof a theorem consists of a series of steps
which, froma set of one or more postulates of the firstgroup as a starting
point, leads us to the theorem, each step being justifiedby an appeal to
one of the rules of procedure.
The postulates of our first group, the rules of procedure, are five in
number:
I. If J is true, if L is wtell-formed,
if all the occurrencesof the variable x
in L are occurrencesas a bound variable, and if the 'cariable y does
not occur ill L, thenK, the result of utbstit'uting
Sx LI for a particular
occurrenceof L in J, is also trite.
II. If J is true,if M and N are ?uell-formed,
if the variableX occursin M,
and if the bound variables in M are distinctbothfrom the variable x
and from thefree variables in N, then K, the result of substituting
Sx MI for a particular occurrenceof {i x. M} (N) in J, is also true.
356
A. CHURCH.
III. If J is true,if M and N are well-formed,

if thevariablex occursin M,
and if the bound variables in M are distinct both from x and from
thefree variables in N, then K, the result of substituting{Ax . M} (N)
for a particular occurrenceof SN MI in J, is also true.
IV. If {F} (A) is true and F and A are well-formed,
then Z (F) is true.
V. If H(F, G) and {F}(A) are true, and F, G, and A are well-formed,
then {G}(A) is true.
And our formal postulates are the thirty-sevenfollowing:
1.
O
JJzp(sr)
(spo qso).
2. 'x3qD
. So(x) (. 1
36
-X(Qf)
3.
(e
4o
)) 3
(x).
* [Qfz W x (z]
so A, *
*
Ye
,
so Ox Y (X Oy )]
Qez
WZ Ox(
[Q
(Pf t)
*
(X))2 .)
* H
o()x
VZ)
(z)
5. 1 (Tf)OT . IH(Tf y,):)? *s (f (x)) O)fxt (f (x)).
'x *So(x)
6. p
.q 11(q,
D *
q(x))
),p
7. So(x,2f (x))
8. (e)
[e W
x).
)+fx
*1(So(x),4 (x))
Q*1ZY [e WZ Ox
o
DSp
9. 'X Sp
24. p
(xp
(X; 2 A
so (X,
A)
O(P
)?tp (x,2f (x))

* [e Wz
Ox
ff(so
Wz 2 0
30
OYU,(y).
qTq(s).pq
(fPxVW)ODq
XOX1
11.19.iso(X
2X) ['0(u,
OTXz
(so)
/)]
.0(u). OH (xy).~0y]9u~('u)
(so W)*ry
12. -X(Sp)3)T Zx (p)(x) .

13.
X(so) 3 T .[so (x) Ox0(x)]
3 tpU(so
14. p Op . q Oq pq.
15. 1q) q vpy.

16. pq Opq q.
17. Zx Ze [T Wx . % 0(x.
I
MO,
18. -'- H(q2 1) Dq),pZ X .

19. Zx o [ 0
0)] O)O n(602

V).
.
.
@(X) H7(IP 0)
(x)
S (?1, x) * 0 (11) * (SO(Y)) 0
@(Y)]0 TU
(SPW))
20. Z)hx. So(x) .

(q)
21. p Op ---q O)
pq.
22.
--I)
p -q Oq -. pq.
23. --p Op ---q Oq'-'p q.

24. p Dv . - .p q] ;)q - q.
V)
.V(u)] O)qou * [[[ap
WX *
(X).
e(XI
x)][a . g
[IH S (X). t(X)]
25.
As.w(it)
1p
26.
27.
tp)
-p
O
p.
28.
T(h) rel (at
29.
2(SP)
ap .
357
POSTULATES
(X)] Ox e (X)] -
(x)JO
D xQ (X)] D
((U).
te) ara
)Obewe( 2 free
).
3(t) 00, l(w,
30. 'X So(X) e me

[n (X)
app
(X)O
n
oU(f,
o)]Ooe
(d))u.
of
31. Sp(l (e)) Air n(O 2 S).

32. E (0~)) Oe Z(0).
) oz S (x te)] [Sp(X
33o [ (x cy)
c r
g(yi t
v)).
Y)
ip (A (Spit)) 00 y(A(Sp,
34. bty
(x, y) pooy ,each) tzSp(i
[Sp(x
y) Oxy 0 (x, y)] Oe
- 0(?t
O Sp(yex)]Oun
ti
And)
(th v)numer
t becounted
(x, y) ity
So(ye,x)]mlpit
v)
Ouv - -V(A (So u)) Op
(A(So, v)).
0 t (A (Sp, v))] O)v pv(it , v)

(A (9, u)) )y
36. E (A (Sp))O~p. Sp(x , y) Oxy~*Sp(y,2) )z Sp(x,2Z)
35.
[it
37. E (A (p)) OT . Sp(x , y) Oxwy

Sp(y, x) .
8. The relation between
free and bound variables.
By a steqp
in a proof we mean an application of one of the rules of procedure IV

or V, occurring in the course of the proof. And in counting the number
of steps in a proof, each step is to be counted with its proper multiplicity.
That is, if a formula M is proved and then used r times as a premise for
subsequent steps of the proof, then each step in the proof of M is to be
counted r times.
If M and N are well-formed and if N can be derived from M by successive applications of the rules of procedure I, II, and III, then M is said
to be convertibleinto N, and the process is spoken of as a conversionof

M into N.
The formulaN is said to be provable as a consequenceof the formulaM,
if M is well-formed, and N could be made a provable formula by adding M
to our list of formal postulates as a thirty-eighthpostulate. Either of the
formulas, M or N, or both, may contain free variables, since although none
of our formal postulates contains free variables, there is, formally, nothing
to prevent our adding a thirty-eighth postulate which does contain free
variables.
We conclude by proving about our system of postulates the three
following theorems:
A. CHURCH.
358
THEOREM
I. Suppose that M containsx as a free variable, and that Xx. M
is provable, and that N is provable as a consequenceof M. Then, if N
containsx as a free variable, theformula [M]Ox N is provable. And if N
does not containx as a free variable, theformula N itself is provable.
We shall prove this theoremby induction with respect to the number
of steps in the proof of N as a consequence of M.
If this numberof steps is zero, then M is convertibleinto N, and, since M
contains x as a free variable, so does N. Hence in this case [M]Ox,.N
may be proved as follows.
Before each formulawe give a symbol which will subsequentlybe used
in referringto it. And after each formula we give the means by which
it is inferred,that is, the numberof the rule of procedureand a referenceto
the premise or premises, or, if the formulais one of our formalpostulates,
the numberof that postulate.
B1:
xx. M
-provable by hypothesis.
A1: {2S 2(q)} (ix . M)
--IIl, B1.
-1.
,Sp)
-V, A2, A1.
As: {2 s (Sp, Sp)} (t x . M)
-IJA3.
A4: [M]OX.M
-by conversion,fromA4.
A5: [Ml]Ox N
A2:
2(Dp)qH(p
Suppose now, as the hypothesisof our induction, that our theorem is

true whenever the number of steps in the proofof N as a consequence of
M is not greater than n. And consider a case in which this number of
steps is n + 1 .
The last step in the proof of N as a consequence of M might be an
application of Rule IV, with premise {F} (A) and conclusion iY(F) or it
mightbe an application of Rule V, with premisesl(F, G) and IF) (A) and
conclusion {G} (A). In the formercase, F might or might not contain x
as a free variable, and A mightor mightnot containx as a free variable.
In the latter case, F, G, and A each mightor mightnot containx as a free
variable. All these possibilities must be considered separately.
Case 1, the last step in the proof of N as a consequence of M is an
applicationof Rule V, A is identicalwith x, and neitherF nor G containsx
as a free variable. We can prove [M]OxDN as follows:
B1:
x. M
Al:
{I(or) 2(7
A3 :
> (ah- [ord (he me

Xb
-provable, by hypothesis.
(2 X. M)
-Illy
IT"(Z Sm
.17()
B1.
.
X/
*0S(X)
AtpRx
(axhM
-V!3,~
359
POSTULATES FOR THE FOUNDATIONOF LOGIC.
-I, As.
A4: {l x [O(x)D)x (x)] Dq *H(Sp, iV)D ,. o(x)DOX (x)}(x . M)
A6: [{ x . M} (x) Ox s (xj]C, .Hs(Sp, i) Op *I{Rx * M} (x) Ox 1 (x) -I, A4.
-II, A5.13
A6: [[M]Ox S (x)] 9q . Il(S, 0p)D . {Ix * M} (x) Ox t (x)
D
-II,
(x)
. [MlOx
A6.
A7: [[M] OX S (x)] DO . n(9, 2V)
induction.
of
by
hypothesis
{
-provable,
F}
(x)
[M]
Ox
B2:
-Ill, B2.
A8: {iS.[M]OxSo(x)}(F)
-V, A7,Ax.
f) D ,.[M] OxVb(x)}(F)
A9:
Sp.17(Sp,
- II, Ag.
Alo: 17(F,2 V) Dip [MlOx V(x)
of induction.
H
by
hypothesis
G)
-provable,
(F,
B8:
-III, B8.
All: {123V (F, tj')}(G)
-V? Alo2All.
A12: 11 J[MIOx V(x)} (G)
-II, A12.
A13: [MlDxfG} (x)
.
N
[M
conversion,fromA18, since {G} (x) is convertiable
-by
Ox
Al4:
into N.

application of Rule V, and F contains x as a free variable, and G and A
do not. We can prove N as follows:
B':
- x. M
provable by hypothesis.
M) -III B1.
A1: le2(e)}(lx.
A2: X(e) De . -NY le) Ox g (x, y )P *[e
-8.
[e(x)Dx e(x, y)]
DOtp(y)
Ox)
IV (SpW
A5:
W] 01 0
(le . 'XY[e(X) OXS(X, A) DI *VX~) 0- H(w(-)
X * M)
- V, AasA.
[QX 3WOxS(X, Oy)]8 (y}I
W
3
* [Q()
* M [(X)
y)]0. s(X,y)] Op
M) x- -I,I(XA8..IIs(X),
(x DOtxt'(y)}(].x.
[e(x)Dx
MI
-2yjl
A6:
-H, A4.
91 tp(y)
[{ x. M}(x)OxS(xy)]
f(0 0)](X),1
M}(x)Dx
y[[M]DOxT(xAy)]Oq.[{x.
A4:
D Mx
O(9(X)
M}(x)
x (x [{U
*
A6: -y[{[x.M}(x)
S y)]A
[JIx . M}(X) Ox (XY Y)] Ov V(y)
Dvt~(y)
[{.x. M}(x)Dxwp(x~y)]
A8:
Xy[[M][DX
y)] D
(x, A
RLMlOx V (X,)]A
[[M]DlO (x
Y q) (Y)
-II,
ip
t)]
Op
- II2 A6.
-II, A8.
(x),
i)]
D ,.
A7.
thatM does not containx as

we may suppose,withoutloss of generality,
"3Evidently,
a boundvariable,because, if the formulaM did containx as a boundvariable,it would
into a formulaH' whichdid not containx as a boundvariable. A siimilar
be convertible
below.
remarkapplies frequently
A. CHURCH.
360
B2:
A9:
-provable, by hypothesisof induction.

[MlOx {F} (A)
-JII, B2.
[M] DX{2x . F} (x, A)
III, A9.
Alo: {fy . [M]Ox { x. F} (x, y)} (A)
All: zy . [M]OX{x.i F} (x, y)
-IV, Alo.
*
A12: {iwSpy *[MlOx Sp(x, y)} (AX F)
III,2 All.x
I I so. [[M]OX11(S (X), V)] 00 -V, A8, A12
2
[[M] Ox S (x, y)] Dy ip(y)} (A *x F)
)] D?. [[M]{ x x.F}
(x,y)]D
y)
A14: [[M]Dxl(Fx.F F}(x),.
Als:
(y)DFx, y]by)
A15: [[M]DOx (F, t)] Di * [[M D {x F}
) -II, A1.
-A1A1.
-II, Al5.
[[MlOx II(F, 2p)] 00 [[MlOxI{F} (Y)] Oyi (Y)
of
induction.
by
hypothesis
-provable,
II(F,
G)
[M]Ox
-III,
B8.
A17: {lIb [M]OxD (F, V))} (G)
A16:
B8:
-V, A16,A17.
A18: {2 t *[[M]Ox {F} (y)] ODVt(y)} (G)
-II, A18.
A19: [IMIOx{F} (y)] Ox {G} (y)
IIJ, B2.
A20: {fy * [M]Dx{F} (y)} (A)
.T,
A21: {2y{G} (y)} (A)
A19,A20.
A22: {G} (A)

N
-II, A21
-by conversion,fromA22.

application of Rule IV, A is identical with x, and F does not contain x
as a free variable. By hypothesis,Xx. M is provable, and from it, by
conversion and an application of Rule IV, we can obtain X . gs(I x. M).
Therefore,using Postulate 9, we can prove Sp(2x . M) OD, (Sp). And hence,
using Postulate 5, we can provef (x, I x . M) OfxZ (f(x)). M~oreover2 x. M
I2 o . S(y)} (x, Ix. M), and this combined
x.
becomes by conversion {xyI
.
(x)).
with the preceding yields {2y2gI .-o(y)} (x, 2x. M)DOx ({2y2gp
Sp(y)}
And this last formulabecomes by conversion[MlDxI{Ix E(x)} (x).
Now using Postulate 3, since Zx. M is provable, we can obtain
Hence, using the last formula
[LM]Ox Sp(x)] D. . H(Sp, i b) D*. [M] Ox f (x).
can
we
in the preceding paragraph,
prove H(I x E(x), p) Dp *[M]Dx iV (x).
And hence, using Postulate 9, we can prove [MIOx. ?s(x) D., (Sp).
Thus we are able to prove [M]OxH (F', G'), where F' stands forI So. SO(x)
and G' for I2p Z (Sp). And, by the hypothesis of our induction, we are
able to prove [M Ox {F'} (A'), where A' stands for F. Therefore, by the
method of Case 2, we can prove {G'} (A'), and fromit by conversion,N.
application of Rule IV, and A contains x as a free variable, and F does
not. Since {F} (A) is convertibleinto {2x{F} (A)} (x), we are able, by the
POSTULATES
361
methodof Case 3, to prove Ix. (F} (A). This last formulais convertible
intoIx. (F} ({(x. A} (x)). Hence, usingPostulate10, we can prove2(F)
N.
and fromit by conversion,
Case 5, the last step in the proof of N as a consequenceof M is an
applicationof Rule IV, and F containsx as a free variable, and A does
not. By the hypothesisof our induction,we can prove [M]Ox {F} (A).
Hence we can proveE(A), and hence, usingPostulate9, s (A)Op (iy).
thisresultwithPostulate5, we can obtainf(x, A) x ~(f(x)).
Andcombining
Now, by the methodofCase 3, we can proveZx. {F} (A) whichis convertibleinto Xx. {Ax. F} (x, A). Combiningthis with the preceding,we
can obtain {Ax. F} (x, A) Ox ((IxF} (x)). And this is convertibleinto
(F} (A)Ox 2 (F).
Thus we are able to proveH(F', G') where F' stands for Axx.{F} (A)
we
and G' standsforIx Z(F). And, by the hypothesisof our induction,
by the methodofCase 1, we
are able to prove[M]Ox{F'} (x). Therefore,
[M]Ox. N.
can prove[M]Ox{G'}(x), and fromit by conversion,
Case 6, the last step in the proofof N as a consequenceof M is an
applicationof Rule IV, and F containsx as a freevariable, and A is
Xx. M is provable,and, by thehypothesis
identicalwithx. By hypothesis,
by the methodof
of our induction,[M]Ox {F} (x) is provable. Therefore,
intoIx. (x
{ . F}(x, x).
Case 3, we can proveXx. { F}(x), and thisis convertible
Z
(F).
{F}
(x)
obtain
can
we
Ox
Hence, usingPostulate11,
Thus we are able to prove H(F', G') where F' stands for Ax. {F} (x)
and G' standsforAx Z (F). And, by the hypothesisof our induction,we
by the methodofCase 1, we
are able to prove [M]Ox {F'} (x). Therefore,
[M]-x. N.
can prove [MlOx{G'}(x), and fromit by conversion,
applicationof Rule IV, and both F and A containx as a free variable.
By the hypothesisof our induction,
[Ml]Ox {F} (A) is provable,and this is
by the method
. A} (y))} (x). Therefore,
{y {F} ({2Ax
into [MlOxD,
convertible
of Case 6, we can prove [MlOxD y. {F} ({Qx. A} (y)).
into [M1OxfIYD2y!(tx. A} (y))}(F).
This last formulais convertible
by the methodof Case 4, we can prove-g' ly -g ({1 x.A} (y)).
Therefore,
thiswithPostulate10,we can obtainXy (jAx.A}(y))] Z(w).
And,combining
withPostulate5, we can obtain
And hence,combining
I y - Uf(x {A
*Ix A) (y{))]OfxI (f(x)).
By themethodofCase 3, using[MlOXIy. {F} ({Ax A} (y)), we can prove
intoXx.Yy.{Ax.F}(x,
xZy-{F}({ x.A} (y)) and thisis convertible
{Ax.A}(y)).
the
preceding
of
end
the
at
formula
the
with
formula
last
this
Combining
24
A. CHURCH.
362
paragraph, we can obtain Xy [{I x . F) (x, {. x . A} (y))] OXi ((2 x . F) (x)).

And this is convertibleinto Xy[{F} ({2x . A) (y)] Ox.2 (F).
Thus we are able to prove fl(F', G') and [MlOx {F'} (x) where F' stands
Therefore,by the
for 2x y. {F} ({2 x . A) (y)) and G' stands for i x (F).
methodof Case 1, we can prove [M]Ox {G') (x), and fromit by conversion,
[MlOx . N.
application of Rule V, and A contains x as a free variable, and F and G
do not. By the hypothesis of our induction, [MlOx{F} (A) is provable.
Hence, by the method of Case 4, we can prove. X(F). And, by the
method of Case 3, we can prove .Yx. {F) (A), which is convertible into
Zx . IF) ({R x .-A}(x)).
Also, by the hypothesisof our induction, H1(F, G) is provable. Hence,
using Postulate 5, we can obtain {F) (f(x)) Ofx{G) (f(x)) and fromthis,using
x . {F)({Ax.A)(x)), we can obtain {F)({AX.A}(x))Ox{G)({ix*A}(x)).
From this last formula,using Rule II, we can obtain lI(F', G'), where F'
and G' stand respectivelyfor Ax. IF) (A) and Ax. {GI (A). And, applying
Rule III to [Ml DX (F) (A), we can prove[M] Ox {F') (x). Therefore,by the
method of Case 1, we can prove [M]Ox {G') (x) and fromit by conversion,
[MlOx. N.
application of Rule V, and G contains x as a free variable, and F and A
do not. By the hypothesisof our induction,{F) (A) is provable, and hence,
using Postulate 2, 1 (F, i) DOpi (A) is provable.
Also, by the hypothesisof our induction,[MI xIH(F, G) is provable, and
fromthis by conversionwe can obtain [M]Ox {2 if'J1(F if'))(G).
Thus we can prove 1l(F', G') and [M]Ox {F') (A') where F' stands for
if'
i 1I(F, f) and G' stands for if'. f'(A), and A' stands for G. Therefore,
by the method of Case 8, we can prove [MlIOx{G') (A') and fiom it by
conversion,[MlOx . N.
application of Rule V, and F and G contain x as a free variable, and A
does not. We can prove [M]Ox. N as follows:
B1: xx . M
-provable,
A,:
{f2e(e)} (Ax *M)
As:
2(e) De *Zy[()
by hypothesis.
-III,
B1.
)]
do.[ e(x)WDX(
DxWOXSo(XIy
,
(M)Y
( D))
ip.
W OxV(X, BY) - 4,
[e z)Ox w(X,BY)]
Oy*e~z
A.,
IIse *-zy [e(x) Ox9w(x, y)] hoT. [e (z) OxII(so (x) +(x))] Do
rem
,,
[eGuOxSO
A O en alW Ox ,
I,]Y))(
BD
-V
A2
Al.
POSTULATES
A4:
{ile.
DX O(X, Y)]
Y[e(T)
363

3
*[Q(X)
DX l(SP(X)
t(X))]
-I, A8.
x.
M} (x) DOxI(p(x), V(x))] 2Xp.
A5: Zy[{IIx. M} (x) OX p(x,y)] OD. [{I
M}
(x) ox (x), y) -]
A: IXy MI Op(xx,y)J
{)x.
A4.
y.
DI
OX ,(SP(x))D ,].
A6: 2y[[M]DlOS(X, y)] OD* [[MI * MI((
H
II, A6.
[{I x . M}I(X)x) 9(X y)A *{i x *M}
(X)MDxI (X Y)
[e(x)
A7:
Dx
(x,
y)A Dy * Q(x)
Dx V(x,
y)} (Ix
'ZY [[M OX SO(x,7y)] DIP . [[M OXn(g (x)

[{ix . MI (x o)x Box AOy) -) .ix.
MI (x
* M)
WAj) DIP
o)x VP(x7 y)
A6.
-11,
A8: Zy[[M] O)xa(x, y)] Op . [[MlOxH(sp(x), V(x))] Op .

[[M]Ox (x, y)AOv. {ix . M} (x)D) tp(x, y)
II, A7.
.
2.
As:
y[[M] ox so(x, y)]
[[MlOx I(p(x), lp(x))J
A
-II, A8.
[[Ml ox s (x, y)]DO. *[MI ox lp (x(, y)
B2: [M]Ox {F} (A)
-provable by hypothesisof induction.
A)
HI7, B.,
t Ox ix y)}
(x.F)
All: fly .[M
VA) IAI,
Ajo.
,ix F} (xIIVA
y)
A12: Xy.[M]lOx
All*
A7: MI H(Fyp[M]Dx))
Dy)(.IxF)
) [I [A12A
{s
.
*
)
OX
Y
A14:
J1((F,{ x.G
[[Ml
}V(x))j -I [[MB]. So(X Y)j
)x t[(x, y) (Fx ()}(F)-V
[M]
AG AllA
I
y)]
F}
A20: [[IMl)Ox[([{A F}(y)
(x)J,V(x))] Dp *[[M]
yOx
{} x *(x)
-Ay
-II A{
[M]o{x (XyY)
)
A16: [[M]Ox {(F,} (x))]
11{2Bx . F} (x y)] Oy
[[MlOx
(II, A15A
[MA OxMD(x {Y)
Ajo: [MlOx {ix
F}
(X,
A17:
[[MlOx
{H(F (x))]
Bs:
[Ml O)x17(F G)
Ox{F} O] Oy[M]Ox (xy)

A[M.l
-provable, by hypothesisof induction.
A18: [MlOx N(F Ix -yGn(x))

Al: cio n [MoRux
l(F, a(x))n (I
A20 :
HI
Bf .
A
G) ad
altr
(IVt* [[Ml Ox {F} (y)] Oy . [Ml Ox V(x y))(Ix.
II A16
ca
G)
- II A20.
All : [[MlOx{F} (y)] Oy . [Ml Ox11x G} (x, y)
A22: [[MlO)xIF) (y)] O)y [MlO)x(GI (y)
-III A2.
A2s: fly *[MlO)xIF) (y)) (A)
-III7 B2.
A,24: (Iy *[MlO)x{G} (y/))
_V, A22, A2g
(A)
II, A24 *
A.,: [MlOx (G) (A)
-by conversion,fromA,^
A26:
[MlO)x .N
_-V, A17
A2 ,

application of Rule V, and F and G and A all three contain x as a free
24*
364
A. CHURCH.
variable. We include, in particular,the case that F and G both containx

as a free variable and A is identical with x, since there is nothing to be
gained by treating this case separately. By the hypothesis of our induction,[Mlcx{F) (A) is provable, and this is convertibleinto
[MlcX{ix {fy. S' F I) (x, {fy. Sx Al) (x))} (x).

Therefore,bythemethodof
Case3, wecanproveXx{Ay. Sx Fj} (x, {2y.SxAj} (x)).
Hence, using Postulate 7, we can prove
{2y.Sx Fj} (x, {2y.SxAl} (x)) x.II({Ay.S
and this is convertibleinto
Flj}(x), V(x)) D) V (x, {2y.S'Al} (x)),
I F) (A) OX. H (F, V(x))
(x, A).
And hence, by the method of Case 1, since [Ml Ox{F} (A) is provable,
we can prove [M]O .Hl(F, b (x)) 2h V(x, A).
Thus we are able to prove[M]:), I(F', G') where F'stands forA4H(F, V(x))
and G' stands for AtV. p(x, A). And, by the hypothesis of our induction,
we are able to prove [M] )D {F'} (A'), where A' stands for Ix. G. Therefore, by the method of Case 10, we can prove [M]2X {G'} (A') and from
it by conversion,[M]OX.N.
application of Rule V, and F and A both contain x as a free variable,
and G does not contain x as a free variable. As before, A may, in
particular, be identical with x. By the hypothesis of our induction,
[M]Ox {F} (A) is provable, and this is convertibleinto [MIOx{ So. So(A)) (F).
Therefore, by the method of Case 7, we can prove [M]OXho . rp(A), and
this is convertible into [M]OX{Ix E(x)} (A). Hence, using Postulate 2,
and the methodof Case 8, we can prove [MlOx.* (A) Do,. H(9, if) Dp V (A).
And hence, by the method of Case 11, since [MlD, {F} (A) is provable,
we can prove [M]9 .l1U(F,DOi
)
V (A).
Thus we are able to prove [MI OXII(F', G') where F' stands for tfH(F, V)
and G' stands for 2 i. iV(A). And, by the hypothesis of our induction,
we are able to prove [M]OX{F') (A') where A' stands for G. Therefore,
by the method of Case 10, we can prove [M]Ox {G'} (A'), and fromit by
conversion,[MlOX. N.
application of Rule V, and A is identical with x, G contains x as a free
variable and F does not. By the hypothesisof our induction,[MI:x (F) (x)
is provable, and this is convertibleinto [MIOx {Ap . 9 (x)} (F). Therefore,
by the method of Case 5, we can prove [M]Ox 2 S.o S (x) and this is con-
POSTULATES
365
vertibleinto [Ml0XE(x). Hence, usingPostulate6, and the methodof

Case 1, we can prove[MlOx. Sp(x)DO . J1(!P,if(x)) DOpt (x, x). Andhence,
by the methodof Case 10, since [M]Ox{F} (x) is provable,we can prove
[MlOx . H(F, V)(x))00 V(x , x).
Thuswe are ableto prove[M]Ox I(F', G') whereF' standsforA lI(F, (x)),
i
and G' standsforA. . V(x,
x). And, by the hypothesisof our induction,
we are able to prove [M]OX{F'} (A') whereA' standsfor Ax. G. Therefore,by the methodof Case 10, we can prove [MlOx {G'} (A') and from
it by conversion,
[M]OX. N.
applicationof Rule V, and G and A both containx as a freevariable,
and F does not containx as a freevariable. By the hypothesisof our
induction,
we can prove [M]OX{F} (A) and hence,by themethodofCase 4,
we can prove2 (F). Hence,usingPostulate5, we can prove l(F, ip)D+p.
{F} (ftx)) DOf' (f(x)). And, by the hypothesisof our induction,we
can prove[Ml:x H(F, G), whichis convertible
into[M]OX{,pIl (F, if)) (G).
Therefore,
by the methodof Case 8, we can prove
[Ml Ox f2o
F)(f (x)) Of
(f (x))
(G),
whichis convertible
into [MlOx. {F} (f(x)) Ofx{G}(f(x)).
Moreover,by conversionfrom[M]Ox {F} (A) we can prove
[MlOx {Ax. {F} ({Ax. A) (x))} (x),
and therefore,
by themethodofCase 3, we can proveXx. {F} ({1 x . A) (x)).
Hence, by the methodof Case 9, using the formulaat the end of the
precedingparagraph,we can prove
[MlDx. {F} ({Ax. A) (x)) Ox {G} ({ix . A) (x)).
Hence,using[M]Ox {l x . {F) ({l x . A) (x))) (x) and the methodof Case 13,
we can prove
[M]Dx{.xe {G)({Ax . A) (x))} (x).
And hencefinally,by conversion,

we can prove[M]OX. N.
applicationof Rule V, and neitherF nor G nor A containsx as a free
variable. By the hypothesisof our induction,1(F, G) and {F) (A) are
bothprovable. Hence, usingRule V, { G) (A) is provable,and fromit by
N.
conversion,
applicationof Rule IV, and neitherF nor A containsx as a freevariable.
A. CHURCH.
366
By the hypothesis of our induction, {F} (A) is provable. Hence, using

Rule IV, Z (F) is provable, and fromit by conversion,N.
COROLLARY.Suppose that M contains x and y as free variablesand that
Z x 2 y . M is provable, and that N is provableas a consequenceof M. Then
if N contains neitherx nor y as a free variable, theformula N is a provable
formula. If N contains y as a free variable, theformula [M]Dxy . N is a
provable formula. And if N contains x as a free variable, theformula
[M]0yx. N is a provableformula.
And similarly for cases where M cotains a greater number of free
variables.
If X (F) is provable, and G does not contain x as a free
THEOREM II.
variable, and {G} (x) is provable as a consequenceof {F} (x), then H(F, G)
is provable.
For, using Postulate 12, we can prove Z x . {F} (x). Therefore, by
TheoremI, {F} (x) Ox {G} (x) is provable. And hence, using Postulate 13,
we can prove 11(F, G).
III. If Z (F) is provable, and N does not contain x as a free
THEOREM
variable, and N is provable as a consequenceof {F} (x), then N is provable.
For, using Postulate 12, we can prove X x. {F} (x). Therefore, by
TheoremI, N is provable.
-Moreoverit is readily seen that if our rules of procedure are to be
and if TheoremsII and III are to be true of our systemthen
Rules I-V
Postulates 1 -13 must be, if not postulates (as we have taken them to
be), at least provable formulas.
PRINCETON UNIVERSITY,
PRINCETON, N. J.

Church - A Set of Postulates For The Foundation of Logic

Încărcat de

Informații document

Titlu original

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

Church - A Set of Postulates For The Foundation of Logic

Încărcat de

Drepturi de autor:

Formate disponibile

Annals of Mathematics

A Set of Postulates for the Foundation of Logic

1. Introduction. In this paper we present a set of postulates for the

in which a, b, and c are used as freevariables, does note state a definite

R(a) R(b)J (c) Dabc . a(b + c) =

FOR THE FOUNDATION OF LOGIC.

A furtherobjection to the use of the free variable is contained in the

in termsof the undefined

FOR THE FOUNDATION OF LOGIC.

We may draw the analogy of a three dimensional geometry used in

to make these presuppositionsas few and as simple as we can, but there

FOR THE FOUNDATION OF LOGIC.

We assume furtherthe meaning and use of the word everyas part of

The expressions { } ( ) and i [ ] are not, of course, single symbols, but

An occurrence of a variable x in a given formulais called an occurrence

FOR THE FOUNDATION OF LOGIC.

I x[M] represents a function,whose value for a value L of the independent

variables. Moreover, alike in the case of functions of one variable and

When F is a single symbol,we shall oftenwrite F x [M] instead of

[iXy[M]]Dx[[M1Dy[N]]we abbreviate furtherto [MDOxy

Moreover,wheneverpossible withoutambiguity,we omitsquare brackets

FOR THE FOUNDATION OF LOGIC.

We introduce at once the followingdefinitions,using an arrow

V(P, Q) is to be read, "P or Q", U(P, Q) is to be read, "P implies Q",

Q (F, G) is to be read "F and G are equivalent" and E(M) is to be read

III. If J is true,if M and N are well-formed,

5. 1 (Tf)OT . IH(Tf y,):)? *s (f (x)) O)fxt (f (x)).

)?tp (x,2f (x))

12. -X(Sp)3)T Zx (p)(x) .

X(so) 3 T .[so (x) Ox0(x)]

15. 1q) q vpy.

18. -'- H(q2 1) Dq),pZ X .

0)] O)O n(602

S (?1, x) * 0 (11) * (SO(Y)) 0

20. Z)hx. So(x) .

23. --p Op ---q Oq'-'p q.

T(h) rel (at

FOR THE FOUNDATION OF LOGIC.

30. 'X So(X) e me

31. Sp(l (e)) Air n(O 2 S).

y) Oxy 0 (x, y)] Oe

Ouv - -V(A (So u)) Op

0 t (A (Sp, v))] O)v pv(it , v)

37. E (A (p)) OT . Sp(x , y) Oxwy

free and bound variables.

in a proof we mean an application of one of the rules of procedure IV

to be convertibleinto N, and the process is spoken of as a conversionof

A1: {2S 2(q)} (ix . M)

Suppose now, as the hypothesisof our induction, that our theorem is

> (ah- [ord (he me

POSTULATES FOR THE FOUNDATIONOF LOGIC.

Case 2, the last step in the proof of N as a consequence of M is an

[JIx . M}(X) Ox (XY Y)] Ov V(y)

thatM does not containx as

-provable, by hypothesisof induction.

A22: {G} (A)

Case 3, the last step in the proof of N as a consequence of M is an

FOR THE FOUNDATION OF LOGIC.

paragraph, we can obtain Xy [{I x . F) (x, {. x . A} (y))] OXi ((2 x . F) (x)).

{f2e(e)} (Ax *M)

FOR THE FOUNDATION OF LOGIC.

'ZY [[M OX SO(x,7y)] DIP . [[M OXn(g (x)

A8: Zy[[M] O)xa(x, y)] Op . [[MlOxH(sp(x), V(x))] Op .